SELECTIVE ATTRITION OF EXPERIMENTAL SUBJECTS - A METHODOLOGICAL STUDY

Rather than being randomly drawn from a well-defined population, experimental samples are frequently formed as the result of a number of nonrandom influences, e.g., volunteering and subjects' attrition. The consequences of one such selective factor were examined in a study that required subjects to report for two experimental sessions. The task involved the multidimensional scaling of complex visual targets, drawings of the human face. The first of two sessions provided data reflecting intrarater consistency, interrater agreement, the orthogonality of perceptual dimensions, and the dimensional complexity of perceptual maps. One group of subjects (n = 23) did not appear for the second session after completing the the first. A second group (n = 23) reported for both sessions. Analysis comparing the two groups indicated that the subjects who did not attend both sessions produced ratings that were significantly lower on measures of both internal consistency and interrater agreement. Perceptual maps obtained from the two groups were structurally different in terms of both the orthogonality and complexity of the perceptual dimensions. The discussion considers the consequences and possible causes of selective attrition.